Astrocytes close a motor circuit critical period

Abstract

Critical periods-brief intervals during which neural circuits can be modified by activity-are necessary for proper neural circuit assembly. Extended critical periods are associated with neurodevelopmental disorders; however, the mechanisms that ensure timely critical period closure remain poorly understood^1,2. Here we define a critical period in a developing Drosophila motor circuit and identify astrocytes as essential for proper critical period termination. During the critical period, changes in activity regulate dendrite length, complexity and connectivity of motor neurons. Astrocytes invaded the neuropil just before critical period closure³, and astrocyte ablation prolonged the critical period. Finally, we used a genetic screen to identify astrocyte-motor neuron signalling pathways that close the critical period, including Neuroligin-Neurexin signalling. Reduced signalling destabilized dendritic microtubules, increased dendrite dynamicity and impaired locomotor behaviour, underscoring the importance of critical period closure. Previous work defined astroglia as regulators of plasticity at individual synapses⁴; we show here that astrocytes also regulate motor circuit critical period closure to ensure proper locomotor behaviour.

Extended Data Figure 1.. Activity-dependent remodeling of motor neuron dendrites during a motor circuit critical period.

(a-e) Tonic activation of motor neurons during embryogenesis induces dendrite retraction. (a) Schematic of the activation paradigm used in this study. For activation of aCC/RP2 motor neurons (MNs), RN2-gal4/lexA drove expression of UAS-CsChrimson::mCherry or lexAop/UAS-CsChrimson::mVenus. Crosses were established on day 0 and fed exclusively on yeast paste supplemented with 0.5 mM all-trans retinal (ATR; required for maximum Chrimson activity) and changed daily for a minimum of 3 days. Timed embryo collections were performed on day 3 for a duration of 1.5 hours (h). Sustained light activation (10550 lx) was followed by immediate dissection. Optogenetic silencing experiments using UAS-GtACR2::EYFP followed the same scheme. (b-e) Activation of aCC/RP2 MNs by Chrimson channelrhodopsin induces dendrite retraction. (b-d) Representative 3D projections of brains expressing Chrimson::mCherry in aCC/RP2 MNs at 0 h after larval hatching (ALH) following activation during embryonic stage 17 (st17). After activation, brains were categorized qualitatively as (b) control, (c) mildly reduced or (d) strongly reduced based on the extent of aCC/RP2 dendritic elaboration (dashed white boxes). Scale, 5 μm. (e) Distribution of each phenotypic class in control, dark-reared animals versus animals whose aCC/RP2 MNs were Chrimson-activated for 15’, 1 h, or 4 h. Dark-reared controls were used throughout as aCC/RP2 MNs show sensitivity to Chrimson in the absence of ATR ((−) ATR) after 15’ and 4 h of Chrimson activation. (f-i) Complementary assays to define the motor circuit critical period. (f-g) Silencing of aCC/RP2 MNs for 1 h by (f) GtACR2 (400 ms pulses of 488 nm light per second) or (g) expression of the temperature sensitive (ts) shibire^ts to block synaptic transmission (active at 30° C), resulted in significant dendrite extension at 0 h ALH, but had no effect at 8 h ALH. N= #animals. 0 h GtACR2: control (N=11), 1 h silencing (N=12). 8 h GtACR2: N=10 per condition. GtACR2 statistics within group (one-way ANOVA): 0 h (p<.0001), 8 h (p<.76). GtACR2 statistics across groups (two-way ANOVA): p<.003. 0 h shibire^ts: control (N=7), 1 h silencing (N=6). 8 h shibire^ts: control (N=6), 1 h silencing (N=7). Shibire^ts statistics within group (one-way ANOVA): 0 h (p<.0002), 8 h (p<.86). Shibire^ts statistics across groups (two-way ANOVA): p<.003. GtACR2 genetics: RN2-gal4, UAS-GtACR2::eYFP. Shibire genetics: RN2-gal4, UAS-shibire^ts, UAS-myr::GFP. (h-i) Activation of aCC/RP2 MNs for 1 h by (h) Chrimson (600 ms pulses of 561 nm light per second) or (i) expression of the thermogenetic activator TrpA1 (inactive at 22° C, fires at ~30 Hz at 27° C), resulted in significant dendrite retraction at 0 h ALH, but had no effect at 8 h ALH. N= #animals. 0 h Chrimson: control (N=12), 1 h activation (N=14). 8 h Chrimson: control (N=12), 1 h activation (N=10). Chrimson statistics within group (one-way ANOVA): 0 h (p<.0001), 8 h (p<.6). Chrimson statistics across groups (two-way ANOVA): p<.001. 0 h TrpA1: control (N=6), 1 h activation (N=11). 8 h TrpA1: control (N=5), 1 h activation (N=6). TrpA1 statistics within group (one-way ANOVA): 0 h (p<.0001), 8 h (p<.25). TrpA1 statistics across groups (two-way ANOVA): p<.0001. Chrimson genetics: RN2-gal4, UAS-Chrimson::mCherry. TrpA1 genetics: RN2-gal4, UAS-TrpA1, UAS-myr::GFP. Throughout: error bars, mean ± SD. Significance: *p<.05; **p<.01; ***p<.001; ****p<.0001; NS, not significant. ♦ used in place of * to denote significance following two-way ANOVA when both one-way and two-way are displayed together. N values reflect biological replicates from 2 technical replicates.

Extended Data Figure 2.. Changes to motor dendrite length and complexity following minutes of altered neuronal activity.

(a-g) Silencing of RP2 by shibire^ts induces dendrite extension. (a-d) MCFO single neuron labeling at 0 h ALH to visualize the morphology of RP2 MN dendrites at 25°C in (a) shibire^ts control (N=43 neurons/N=24 brains) compared to neurons silenced with shibire^ts to block synaptic transmission (active at 30° C) for (b) 15 minutes (N=18/N=15), (c) 1 h (N=7/N=6), or (d) 3 h (N=29/N=18). Prime panels show reconstructions of RP2 dendritic arbors (performed using the Imaris “Filaments” tool). Blue dots show the seed positions for each Filament. Scale, 5 μm. Genetics: RN2-gal4,UAS-shibire^ts,UAS-hsMCFO. (e-h) Quantification of (e) total dendrite length, (f) longest branch length (measure of distal dendrite extension), (g) # of dendritic branch points, and (h) the distribution of dendrite lengths per reconstructed neuron (% of all processes) post-silencing by shibire^ts. Statistics (one-way ANOVA) by increasing length of silencing: (e) p<.05, p<.0001, p<.0001; (f) p<.007, p<.02, p<.0001; (g) p<.54, p<.14, p<.8; (h) the percentage of long processes (>2 μm) were significantly increased after 1 h (p<.0001) and 3 h (p<.0001) of silencing (subtle decreases after 15’, p<.04). (i-l) Remodeling of dynamic distal processes within 12 minutes of aCC/RP2 activation. (i) Schematic depicting a larval brain at 0 h ALH with aCC/RP2 MNs in purple. Two hemisegments were imaged per experiment (box). (j-k) MN Chrimson activation results in dendrite retraction within minutes. (j) 3D projection of a control isolated CNS at 0 h ALH, time 0 (RN2-gal4,UAS-myr::GFP; + ATR). Yellow box highlights intersegmental region used for reconstruction of individual dendrites. Scale, 5 μm. (j’-j”) 3D projections from representative time points over a 15-minute acquisition period. Left panels, myr::GFP signal alone. Yellow arrowheads mark the tip of a single reconstructed process. Right panels, green Imaris “Filament” reconstruction of indicated process. Scales, 1 μm. (k) 3D projection of an isolated CNS at 0 h ALH for Chrimson-activation, time 0 (RN2-gal4,UAS-CsChrimson::mVenus; + ATR). Yellow box highlights intersegmental region used for reconstruction of individual dendrites. Scale, 4 μm. (k’-k”) 3D projections from representative time points over a 15-minute acquisition period. Left panels, Chrimson::mVenus signal alone. Yellow arrowheads mark the tip of a single reconstructed process. Right panels, green Imaris “Filament” reconstruction of indicated process. Scales, 1 μm. (l) Quantification (one-way ANOVA) of normalized dendrite length over time in myr::GFP controls versus brains that were Chrimson-activated for 3 min (p<.99), 8 min (p<.06), 12 min (p<.05), or 15 min (p<.02). N=10 processes each from N=4 brains per condition, with processes binned by length into 10 categories. Control length remained stable over the 15-minute acquisition period. Chrimson-activation results in progressive retraction of MN dendrites. Control box plot specifications (min, max, centre, upper box bound (75%), lower box bound (25%), minus whisker, plus whisker): 0 min (−.89, 1.22, .51, .65, −.13, −.76, .58), 3 min (−.96, 1.78, .40, .95, −.09, −.87, .82), 8 min (−1.57, 1.33, .23, .92, −.77, −.80, .42), 12 min (−1.92, 1.48, .30, 1.05, −.40, −1.51, .43), 15 min (−.89, 1.80, .18, .97, −.62, −.27, .84). Chrimson box plot specifications: 0 min (−.89, 1.22, .51, .65, −.13, −.76, .58), 3 min (−.46, 1.74, .10, .95, −.29, −.17, .79), 8 min (−1.25, 1.14, −.01, .88, −.83, −.42, .26), 12 min (−1.52, 1.01, −.67, .52, −1.35, −.18, .49), 15 min (−1.79, .99, −.95, .34, −1.45, −.34, .65). Throughout: error bars, mean ± SD. Significance: *p<.05; **p<.01; ***p<.001; ****p<.0001; NS, not significant. N values reflect biological replicates from 2 technical replicates.

Extended Data Figure 3.. Quantification of the number of synaptic connections between the GABAergic A23a or cholinergic A18b interneurons and the motor neurons aCC/RP2.

(a) TEM reconstruction of the A23a premotor neuron in a first instar larval brain at 4 h ALH; pre-synapses are primarily localized to the contralateral branch (arrow). (b) Light microscopy image of a single A23a premotor neuron at 4 h ALH (78F07-lexA) with cyan membranes (lexAop-myr::GFP) and magenta pre-synapses (lexAop-brp-short::cherry). Most synapses with aCC/RN2 MNs are at the contralateral process (arrow). Note the morphological similarity between light and EM images of A23a. Asterisks, sparse off-target expression not in A23a. Scale, 2 μm. (c-d) A23a is GABAergic. Scale, 3 μm. Genotype: 78F07-lexA lexAop-myr::GFP. (e) Representative image of A23a (A1L) forming 14 synapses (white dots) with aCC (A1R) in the TEM reconstruction. Dorsal view, midline to left. Quantification of A23a-aCC synapses from the TEM reconstruction: 21 in A1R and 13 in A1L; A23a-RP2 synapses: 2 in A1R and 3 in A1L. (f) Representative image of A18b (A1L) forming 11 synapses (white dots) with aCC (A1R) in the TEM reconstruction. Dorsal view, midline to left. Quantification of A18b-aCC synapses from the TEM reconstruction: 18 in A1R and 10 in A1L; A18b-RP2 synapses: 7 in A1R and 5 in A1L. (g-g””) Quantification of “putative” A23a-aCC/RP2 synapses by light microscopy at 4 h ALH. Scales, 2 μm. (g) Representative 3D projection of aCC/RP2 dendrite membrane (Chrimson::mVenus+; magenta) and A23a Brp-short puncta (white). Genotype: RN2-gal4,UAS-Chrimson::mVenus x 78F07-lexA,lexAop-brp-short::cherry. (g’) aCC/RP2 dendrite membrane (Chrimson::mVenus+); (g”) Imaris “Surface” rendering of g’; (g’”) A23a pre-synapses (Brp-short::cherry+); (g””) Imaris “Spots” measurement of Brp-short puncta within 90 nm of dendritic membrane (red dots, 19 putative direct synapses).

Extended Data Figure 4.. Remodeling of MN synapses during and after critical period closure.

(a-f’) Remodeling of pre-synapses during the critical period. (a-c) Imaris “Surface” from (a) control or post-Chrimson activation for (b) 1 h or (c) 4 h, terminating at 4 h ALH (critical period open). Magenta, dendrite marker. White, presynaptic Brp-short::Cherry puncta from the excitatory A18b neuron. (a’-c’) Imaris “Spots”, presynaptic Brp puncta within 90 nm of dendritic surface. Scale, 2 μm. Genotype: RN2-gal4,UAS-Chrimson::mVenus; 94E10-lexA,lexAop-brp-short::cherry. (d-f) Imaris “Surface” from (d) control or post-Chrimson activation for (e) 1 h or (f) 4 h, terminating at 4 h ALH (critical period open). Magenta, dendrite marker. White, presynaptic Brp-short::Cherry puncta from the excitatory A23a neuron; (d’-f’) Imaris “Spots”, presynaptic Brp puncta within 90 nm of dendritic surface. Scale, 2 μm. Genotype: RN2-gal4,UAS-GtACR2::eYFP; 78F07-lexA,lexAop-brp-short::cherry. Activation caused decreased numbers of excitatory, but not inhibitory, synapses (quantified in Fig. 2t). Overall, we observed Brp puncta numbers matching synapse numbers by TEM in stage-matched control brains (4 h ALH; A18b: 19.5±4.9 Brp+ puncta vs. 20±2.5 TEM synapses per hemisegment; A23a: 16.9±4.1 vs. 19.5±3.5, see Fig. 2 legend for Ns). (g-l) Stability of pre-synapses after critical period closure. (g-h) Imaris “Surface” from (g) control or (h) post-Chrimson activation from 7–8 h ALH (critical period closed; magenta, dendrite marker) with presynaptic Brp-short::Cherry puncta (white) from the excitatory A18b neuron; (g’-h’) Imaris “Spots”, presynaptic Brp puncta within 90 nm of dendritic surface. Scale, 2 μm. Genotype: RN2-gal4,UAS-Chrimson::mVenus; 94E10-lexA,lexAop-brp-short::cherry. (i-j) Imaris “Surface” from (i) control or (j) post-GtACR2 silencing from 7–8 h ALH (critical period closed; magenta, dendrite marker) with presynaptic Brp-short::Cherry puncta (white) from the inhibitory A23a neuron; (i’-j’) Imaris “Spots”, presynaptic Brp puncta within 90 nm of dendritic surface. Scale, 2 μm. Genotype: RN2-gal4,UAS-GtACR2::eYFP; 78F07-lexA,lexAop-brp-short::cherry. (k-l) Quantification (one-way ANOVA) of synapse number following (k) MN excitation or (l) inhibition. N = #hemisegments/#animals: A18b Chrimson N= 10/8 (control); 15/9 (1 h activation, p<.57). A18b GtACR2 N= 28/10 (control); 22/9 (1 h silencing, p<.94). A23a Chrimson N= 24/14 (control); 20/14 (1 h activation, p<.63). A23a GtACR2 N=22/10 (control); 25/10 (1 h silencing, p<.52). (m-t) Remodeling of excitatory post-synaptic densities during and after the critical period. (m-o) Representative 3D projection showing dendrite membranes (magenta) and post-synaptic densities (green) in (m) control or following (n) 1 h or (o) 4 h MN activation. (m’-o’) Imaris “Spots”, post-synaptic puncta within 70 nm of dendritic surface. Scale, 2 μm. Genotype: RN2-lexA,lexAop-Chrimson::tdTomato, lexAop-drep-2::GFP. (p-r) Representative 3D projection showing dendrite membranes (magenta) and post-synaptic densities (green) in (p) control or following (q) 1 h or (r) 4 h MN silencing. (p’-r’) Imaris “Spots”, post-synaptic puncta within 70 nm of dendritic surface. Scale, 2 μm. Genotype: RN2-gal4,UAS-GtACR2::eYFP, UAS-drep-2::mStrawberry. (s-t) N= #animals, synapse # averaged across 4 hemisegments (A1-A2). N=10 per all conditions and controls. (s) Quantification (one-way ANOVA) of excitatory post-synapse number following MN excitation for 1 h (p<.0002) or 4 h (p<.0001) relative to control, and following inhibition for 1 h (p<.4) or 4 h (p<.005) relative to control at 4 h ALH. (t) Quantification (one-way ANOVA) of excitatory post-synapse number following MN excitation (p<.9) or silencing (p<.49) for 1 h relative to control at 8 h ALH. Throughout: Error bars, normalized mean (0) ± SEM. Significance: *p<.05; **p<.01; ***p<.001; ****p<.0001. N values reflect biological replicates from 2 technical replicates.

Extended Data Figure 5.. Progressive ensheathment of motor synapses by astrocytes across critical period closure.

(a-e) Time course of astrocyte-MN synapse association from (a) embryonic stage 17, (b) 0 h ALH, (c) 4 h ALH, (d) 8 h ALH, and (e) 22 h ALH. Astrocytes, cyan (Gat+). MN membranes, green (myr::GFP+). MN post-synapses, magenta (mStrawberry+). (a’-e’) Astrocytes and post-synapses alone. Synapses ≤ 90 nm from astrocyte membranes were counted as ensheathed. Scale, 5 μm. Genotype: RN2-gal4, UAS-myr::GFP, UAS-drep-2::mStrawberry. (f) Quantification of astrocyte-associated post-synapses (% of total) revealed a significant interaction between developmental stage and % ensheathment (two-way ANOVA, p<.0001). N= 6 brains per timepoint, % ensheathment averaged over N≥2 segments (A1-A2). Error bars, mean ± SD. Significance: ****, p<.0001. N values reflect biological replicates from 2 technical replicates.

Extended Data Figure 6.. Astrocyte ablation and manipulation extends critical period plasticity.

(a-i) Astrocytes close the critical period. (a-c) Representative 3D projections of brains expressing Chrimson::mCherry in aCC/RP2 motor neurons (RN2-gal4,UAS-Chrimson::mCherry) illustrating the three classes of dendritic arbor morphology at 8 h ALH following 4 h of Chrimson activation: (a) control, (b) mildly reduced, and (c) strongly reduced dendritic arbor size/complexity. Scale, 10 μm. (d) Quantification of each phenotypic class. Denoted N= #animals. Control animals show no significant dendritic remodeling after 15’ of activation at this stage (p<.12, one-way ANOVA). In contrast, ablation (abl.) of astrocytes results in a significant shift in the distribution of phenotypic classes away from wildtype (no light abl. versus 15’ activation abl., p<.03, one-way ANOVA). Loss of astrocytes strongly sensitized these motor neurons to remodeling (p<.04, two-way ANOVA). Note that control and 4 h data are also displayed in Fig. 3e. Control Genotype: RN2-gal4,UAS-Chrimson::mCherry; alrm-lexA,lexAop-myr::GFP. Ablation Genotype: RN2-gal4,UAS-Chrimson::mCherry; alrm-lexA,lexAop-rpr. (e-h) Representative 3D projections of aCC/RP2 dendrites at 8 h ALH. Scale, 5 μm. (e-f) Dark-reared controls with (N=13) or without (N=15) astrocyte ablation. (g-h) GtACR2 silencing in aCC/RP2 from 7–8 h ALH with (N=12) or without (N=12) astrocyte ablation; note that astrocyte ablation prolongs the critical period to allow activity-dependent dendrite extension. N= #animals, volume averaged over 4 independent hemisegments (A1-A2). Genotypes: RN2-gal4,UAS-GtACR2::eYFP; alrm-lexA (control), RN2-gal4,UAS-GtACR2::eYFP; alrm-lexA,lexAop-rpr (ablation). (i) Quantification by two-way ANOVA (p<.009). (j-n) Astrocytes do not dampen critical period plasticity. (j-m) Representative 3D projections of aCC/RP2 dendrites at 0 h ALH. Scale, 5 μm. (j-k) Dark-reared controls with (N=5) or without (N=7) astrocyte ablation. (l-m) Chrimson activation in aCC/RP2 for 1 h in late embryo terminating at 0 h ALH, with (N=7) and without (N=6) astrocyte ablation; note that astrocyte ablation does not enhance activity-induced dendrite retraction. N= # animals, volume averaged over 4 independent hemisegments (A1-A2). Control Genotype: RN2-gal4,UAS-Chrimson::mCherry; alrm-lexA,lexAop-myr::GFP. Ablation Genotype: RN2-gal4,UAS-Chrimson::mCherry; alrm-lexA,lexAop-rpr. (n) Quantification by two-way ANOVA (p<.74). (o-p’) Representative images of astrocyte morphology in (o-o’) control or following (p-p’) astrocyte ablation. White, astrocyte membranes (Gat+). Anterior to the left, dorsal is up. Scale, 10 μm. Control Genotype: RN2-gal4,UAS-Chrimson::mCherry; alrm-lexA,lexAop-myr::GFP. Ablation Genotype: RN2-gal4,UAS-Chrimson::mCherry; alrm-lexA,lexAop-rpr. MN channel not shown. (q-v) MCFO clones showing single astrocyte morphology and volume in control (N=38/13) or following KD of gat (N=11/7), chpf (N=12/4), nlg4 (N=23/10), or nlg2 (N=23/7) at 8 h ALH. N= #clones/#animals. The pan-astrocyte marker Gat was used to assay astrocyte ablation at 8 h ALH. Scales, 5 μm. Normalized, mean astrocyte volume at the bottom of each MCFO panel (via Imaris “Surface”). Statistics (one-way ANOVA) relative to control: gat (p<.0001), chpf (p<.43), nlg4 (p<.007), nlg2 (p<.37) denoted by asterisks. Genotype: alrm-gal4,UAS-hsMCFP,UAS-RNAi. (w-x’) Representative images showing labeling of all astrocytes by MCFO in (w-w’) control or (x-x’) following astrocyte KD of nlg2. Anterior to the top, dorsal is up. Astrocytes tile the entire the CNS and exhibit normal tiling behavior, as exhibited by non-overlapping territories in single z-slices). Scales, 8 μm. Genotype: alrm-gal4,UAS-hsMCFP,UAS-RNAi. Throughout: error bars, mean ± SD. Significance: *p<.05; **p<.01; ***p<.001; ****p<.0001; NS not significant. ♦ used in place of * to denote significance following two-way ANOVA when both one-way and two-way are displayed together. N values reflect biological replicates from 2 technical replicates.

Extended Data Figure 7.. Expression of Nrx-1 in motor dendrites during the critical period.

(a-b’”) Localization of Nrx-1 (magenta) relative to MN dendritic membranes (orange, myr::GFP+) and pre-synapses (green, Brp+) in control (N=15) and following MN-specific KD of nrx-1 at 0 h ALH (N=20). N= #animals. (a-a’”) Nrx-1 colocalized with motor dendrite membranes (white circle in inset) and synapses (Brp+) in control. (b-b’”) Nrx-1 colocalized with synapses, but was absent from dendritic membranes in KD brains, as evidenced by increased clarity in Nrx-1+ synaptic puncta (white circle in inset). Scale, 2 μm. Insets, 1.5X zoom. Genotype: RN2-gal4, UAS-myr::GFP, UAS-RNAi or control. (c-c’) Representative image showing localization of a Nrx-1::GFP fusion (magenta) relative to MN dendrites (green, Denmark::cherry+). Nrx-1::GFP is present within motor dendrites (white circle, colocalization channel in c’”). N= 15 animals. Scale, 2 μm. Genotype: RN2-gal4, UAS-Nrx-1::GFP, UAS-Denmark::cherry. N values reflect biological replicates from 2 technical replicates.

Extended Data Figure 8.. Expression of Nlg2 in astrocytes during the critical period is not required for proper E/I synapse balance.

(a-b’”) Localization of Nlg2 (magenta) relative to astrocyte membranes (green, mcd8GFP+) and pre-synapses (blue, Brp+) in control (N=15) and following astrocyte-specific KD of nlg2 at 0 h ALH (N=18). N= #animals. (a-a’”) Nlg2 colocalized with astrocyte membranes (mcd8GFP+) and synapses (Brp+) in control. (b-b’”) Nlg2 colocalized with synapses (yellow box), but was absent from astrocyte membranes in KD brains, as evidenced by increased clarity in Nlg2+ synaptic puncta. Scale, 2 μm. Insets, 1.5X zoom. Genotype: alrm-gal4, UAS-mcd8::GFP, UAS-RNAi or control. (c-c’”) Representative images showing aCC/RP2 dendrites (magenta) and excitatory post-synapses (cyan, Drep-2::GFP+) relative to all presynapses (orange, Brp+) in control (N=7) versus (d-d’”) astrocyte-specific KD of nlg2 (N=9). (c’”, d’”) Note close apposition of pre- and post-synapses (arrowheads). N= #animals, with synapses averaged over 4 hemisegments per brain (A1-A2). Scale, 3 μm. Genotype: RN2-lexA, lexAop-myr::tdTomato, lexAop-drep-2::GFP; alrm-gal4, UAS-RNAi or control. (e-f) Quantification of (e) total synapse number (p<.11) and (f) the ratio of excitatory synapses to total synapses (p<.96) revealed no significant differences (Mann-Whitney test, two-sided). Error bars, mean ± SD. NS, not significant. N values reflect biological replicates from 2 technical replicates.

Extended Data Figure 9.. Tubulin stability correlates with dendrite retention during activity-induced remodeling.

(a-e) Dendrites with stable microtubules are resistant to activity-induced remodeling. Representative 3D projections of brains expressing Chrimson (green) and the microtubule reporter Zeus (a tagged microtubule binding protein, magenta) in aCC/RP2 MNs (RN2-gal4,UAS-Cherry::Zeus,UAS-CsChrimson::mVenus) at 0 h ALH. Brains were preserved with cold fixative to visualize stable microtubule populations in (a) control and after Chrimson-activation for (b) 15’, (c) 1 h, or (d) 4 h. Prime panels show Cherry:Zeus channel alone. Scale, 10 μm. Boxed in regions represent regions of interest (ROIs) that were used for Imaris “Surface” reconstructions to determine dendrite and microtubule volume. (e) Quantification (one-way ANOVA) of the normalized volume of dendrite membranes (Chrimson::mVenus+) and Cherry::Zeus within the same ROI. Microtubule volumes at each time point were calculated relative to the membrane volume for dark-reared controls. N= # animals, with the volume per animal representing the average volume across 4 hemisegments (A1-A2). In dark-reared controls (N=4), stable microtubule populations reflect 55±8% of the total dendritic volume. Chrimson-activation results in a significant decrease in total dendritic volume after 15’ (N=6, p<.05) and 1 h (N=4, p<.0003) of activation. Microtubule volume is unchanged after 15’ (p<.26) or 1 h (p<.35). After 4 h of activation (N=6), both membrane volume (p<.0001) and microtubule volume (p<.0002) are significantly reduced; however, dendrites with stable microtubules are preferentially retained such that membrane volume is nearly equivalent to the Cherry::Zeus volume (#, p<.02). Error bars, mean ± SD. Significance: *p<.05; **p<.01; ***p<.001; ****p<.0001. N values reflect biological replicates from 2 technical replicates.

Extended Data 10.. Assaying permanent motor circuit changes following manipulation of critical period activity.

(a-c) Stability of remodeled synapses following MN activation during the critical period (CP). (a-b) Imaris “Surface” from (a) control or (b) post-Chrimson activation from 0–4 h ALH (critical period open; magenta, dendrite marker) followed by 20 h recovery in dark conditions. Presynaptic Brp-short::Cherry puncta (white) from the excitatory A18b neuron; (a’-b’) Imaris “Spots”, presynaptic Brp puncta within 90 nm of dendritic surface. Scale, 2 μm. Genotype: RN2-gal4,UAS-Chrimson::mVenus; 94E10-lexA,lexAop-brp-short::cherry. Data not shown: inhibitory A23a synapses following critical period activation and recovery. Genotype: RN2-gal4,UAS-Chrimson::mVenus; 78F07-lexA,lexAop-brp-short::cherry. (c) Quantification of excitatory (p<.002) and inhibitory (raw images not shown; p<.61) synapse numbers following MN excitation by one-way ANOVA. N = #hemisegments/#animals: A18b Chrimson N= 19/3 (control); 15/2 (1 and 4 h activation combined). A23a Chrimson N= 15/2 (control); 31/8 (1 and 4 h activation combined). Error bars, normalized mean (0) ± SEM. (d-g’) Validation of genetic tool for conditional KD of astrocyte genes to transiently extend the critical period. N=10 animals per condition. Orthogonal views through the ventral nerve cord showing the extent of astrocyte infiltration (Gat+, green) into the synapse-dense neuropil (Brp+, magenta). Prime panels show Gat signal alone. (d-e’) In control animals (25H07-gal4 X UAS-myr::GFP), astrocytes progressively infiltrate the neuropil from 8 h ALH through 44 h ALH. (f-f’) When reared at 30°C through 8 h ALH, expression of UAS-htl^DN in astrocytes (tubP-gal80^ts; 25H07-gal4) suppressed astrocyte infiltration. (g-g’) Shifting to 18°C at 8 h ALH resulted in inhibition of Gal4 by TubP-Gal80^ts, reduced expression of htl^DN, and rescued astrocyte infiltration at 44 h ALH (25°C standard, see methods for details on staging). 8 h scale, 20 μm. 44 h scale, 30 μm. Significance: **, p<.01. N values reflect biological replicates from 2 technical replicates.

Figure 1.. A critical period for motor circuit plasticity.

(a) Schematic for reader orientation. A, anterior. P, posterior. L, left. R, right. CNS, central nervous system. MNs, motor neurons. (b-m) aCC/RP2 dendrites (single hemisegment) from (b-e) dark-reared control, and following optogenetic (f-i) silencing or (j-m) activation for 1 h ending at the indicated stage. Per stage: left, full dendritic arbor; right, RP2 clone. Scales, 5 μm. Genotypes: Control and silencing: RN2-gal4,UAS-GtACR2::EYFP; activation: RN2-gal4,UAS-CsChrimson::mCherry. RP2 clones: + UAS-hsMCFO. (n) Quantification, full arbor. N= #animals, ordered by increasing stage. Silencing: controls: N=8, 13, 14, 11; experimentals: N=7, 16, 25, 21. Silencing statistics by increasing stage: within group (one-way ANOVA): p<.007, p<.01, p<.28, p<.66; across groups (two-way ANOVA): p<.0001. Activation controls and experimentals: N= 6 per condition/stage. Activation statistics by increasing stage: within group (one-way ANOVA): p<6.3×10⁶, p<.003, p<.07, p<.46; across groups (two-way ANOVA): p<.0001. (o-p) Quantification, clones. N = #neurons/#animals by increasing stage. Silencing: controls: N=19/15, 13/10; experimentals: N=14/13, 11/10. Silencing statistics (two-way ANOVA): p<.004. Activation: controls: N=7/6, 17/14; experimentals: N=17/17, 8/8. Activation statistics (two-way ANOVA): p<.008. (q-t) aCC/RP2 dendritic arbor following embryonic activation (st17) and recovery by dark-rearing (0 h vs. 22 h ALH). Scale, 5 μm. Genotype: RN2-gal4,UAS-CsChrimson::mCherry. (u) Quantification. N= # animals, ordered by increasing stage. Controls: N=5, 10. Experimentals: N=6, 9. Statistics (two-way ANOVA): p<.0001. Throughout: Error bars, mean ± SD. Significance: **p<.01; ****p<.0001; NS= not significant. ♦ significance following two-way ANOVA. Ns: biological replicates from 3 technical replicates.

Figure 2.. Activity-dependent scaling of dendrite length and synaptic inputs.

(a-f) GtACR2 silencing (or control) for the indicated times prior to 0 h ALH. RP2 clones: MCFO. N = #neurons/#animals: N=21/15, 12/9, 13/9, 15/9, respectively. Scale, 5 μm. Genotype: RN2-gal4,UAS-GtACR2::EYFP,UAS-hsMCFO. (e-f) Quantification of morphology. Error bars, mean ± SD. Statistics (one-way ANOVA) by increasing length of silencing for (e) p<.46, p<.02, p<.005; (f) p<.18, p<.02, p<.008. (g-l) Chrimson activation (or control) for the indicated times prior to 0 h ALH. RP2 clones: MCFO. N=18/15, 7/6, 16/11, 29/19, respectively. Scale, 5 μm. (a-l) Genotypes: RN2-gal4,UAS-CsChrimson::mCherry,UAS-hsMCFO. (k-l) Quantification of morphology. Error bars, mean ± SD. Statistics (one-way ANOVA) by increasing length of activation for (k) p<.002, p<.001, p<.0001; (l) p<.02, p<.0001, p<.0001. (m-o) Imaris “Surface” rendering of dendritic membranes (magenta) in control or post-GtACR2 silencing terminating at 4 h ALH. White, presynaptic puncta, A23a neuron (inhibitory). Scale, 2 μm. (m’-o’) Imaris “Spots” rendering of puncta ≤ 90 nm of dendritic surface. Genotype: RN2-gal4,UAS-GtACR2::eYFP; 78F07-lexA,lexAop-brp-short::cherry. (p-r) Imaris “Surface” rendering of MN dendrites (magenta) from (p) control or post-GtACR2 silencing terminating at 4 h ALH. White, presynaptic puncta, A18b neuron (excitatory); (p’-r’) Imaris “Spots” rendering of puncta ≤ 90 nm of dendritic surface. Scale, 2 μm. Genotype: RN2-gal4,UAS-GtACR2::eYFP; 94E10-lexA,lexAop-brp-short::cherry. (s-t) Quantification (one-way ANOVA) following MN (s) inhibition or (t) excitation. N = #hemisegments/#animals: A23a GtACR2 N= 71/23 (control); 36/10 (15’ silencing, p<.72); 47/17 (1 h silencing, p<.0001); 22/10 (4 h silencing, p<.0001). A18b GtACR2 N= 44/19 (control); 17/9 (15’ silencing, p<.46); 17/8 (1 h silencing, p<.22); 17/9 (4 h silencing, p<.0001). A23a Chrimson N= 33/11 (control); 30/9 (15’ activation, p<.38); 22/5 (1 h activation, p<.13); 29/7 (4 h activation, p<.14). A18b Chrimson N= 18/6 (control); 19/8 (15’ activation, p<.1); 21/6 (1 h activation, p<.0001); 23/10 (4 h activation, p<.0001). Error bars, normalized mean (0) ± SEM. Throughout: Significance: *p<.05; **p<.01; ****p<.0001; NS= not significant. Ns: biological replicates from 3 technical replicates.

Figure 3.. Astrocytes terminate the critical period.

(a-b) aCC/RP2 dendrites ± astrocyte ablation in controls or (c-d) following Chrimson activation (4–8 h ALH). Scale, 5 μm. Genotypes: RN2-gal4,UAS-CsChrimson::mCherry; alrm-lexA + lexAop-myr::GFP (control) or lexAop-rpr (ablation). (e) Quantification. Denoted N= #animals. Control statistics (one-way ANOVA): p<.72. Activation statistics (one-way ANOVA): p<.0001. Statistics across groups (two-way ANOVA): p<.04. (f-h) Live imaging: dendrite dynamics. 3D projection, one hemisegment of aCC/RP2 dendrites, 0 h ALH. Yellow boxes, reconstructed regions. Scale: 5 μm. (g’-h’) Dynamic dendrite filopodia (arrowheads) over time. Scale, 1 μm. Genotypes: Control: RN2-gal4,UAS-myr::GFP; alrm-lexA. Ablation: +lexAop-rpr. (i-k) Quantification. N=50 dendrites from 5 animals per timepoint. (i) Dendrite displacement statistics across stages (two-way ANOVA): p<.0001. Statistics (one-way ANOVA) comparing 22 h ablation with 4 h (p<.71), 8 h (p<.16) and 22 h (p<.0001) controls. (j) Statistics (Fisher’s exact test, two-sided) assaying distribution of dendrite dynamics at 4 h (p<.08), 8 h (p<.04), and 22 h (p<.0001) relative to 0 h control, or relative to 22 h ablation (0h: p<.007; 4 h: p<.007; 8 h: p<.91, 22 h: p<.0001). (k) Statistics (one-way ANOVA) to assay extension at (4 h: p<.81; 8 h: p<.02; 22 h: p<.01) or retraction (4 h: p<.99; 8 h: p<.76; 22 h: p<.01) relative to 0 h control. Statistics to assay control extension (0 h: p<.04; 4 h: p<.2; 8 h: p<.27; 22 h: p<.01) or retraction (0 h: p<.67; 4 h: p<.9; 8 h: p<.99; 22 h: p<.001) relative to 22 h ablation. Throughout: Error bars, mean ± SD. Significance: *p<.05; **p<.01; ****p<.0001; NS= not significant. ♦ significance following two-way ANOVA. ψ: significance for Fisher’s exact tests. Ns: biological replicates from 3 technical replicates.

Figure 4.. Neuroligin to Neurexin signaling stabilizes microtubules and closes the critical period.

(a-k) Factors in astrocytes (a-f) or MNs (h-j) required to close the critical period by 8 h ALH. Top row, dark-reared controls; bottom row, experimentals. Scales, 5 μm. Genotypes: (a,c-f) lexAop-CsChrimson::mVenus,RN2-lexA,alrm-gal4 x UAS-RNAi or control, (b) RN2-gal4,UAS-Chrimson::mCherry,lexAop-rpr; alrm-lexA (h-j), RN2-gal4,UAS-CsChrimson::mCherry x UAS-RNAi or control. (g,k) Quantification. Denoted N= #animals. (g) Astrocyte statistics (two-way ANOVA): ablation (p<.005), gat KD (p<.05), chpf KD1 (p<.0001), chpf KD2 (o<.0001), nlg4 KD1 (p<.002), nlg4 KD2 (p<.0002), nlg2 KD1 (p<.006), nlg2 KD2 (p<.03). (k) MN statistics (two-way ANOVA): lar KD1 (p<.001), lar KD2 (p<.002), nrx-1 KD1 (p<.003), nrx-1 KD2 (p=.05). (l-o) Overexpression of (m) Nrx-1 in MNs or (n) Nlg2 in astrocytes closes the critical period, 4 h ALH. Scale, 5 μm. Genotypes: (l-m) RN2-gal4,UAS-CsChrimson::mCherry x UAS-myr::GFP or UAS-Nrx-1, (n) RN2-lexA,lexAop-Chrimson::tdTomato,alrm-gal4 x UAS-Nlg2, or UAS-myr::GFP control (not shown). (o) Quantification. Denoted N= #animals. Statistics within group (one-way ANOVA): MN control (RN2>GFP, p<.0001), astrocyte control (alrm>GFP, p<.0001), Nrx-1 OE (p<.14), Nlg2 OE (p<.0002). Statistics across groups (two-way ANOVA): Nrx-1 OE (p<.005), Nlg2 OE (p<.0007). (p) Live imaging of aCC/RP2: Chrimson::mVenus (green membranes) and Cherry::zeus (stable microtubules, heatmap), 0 h ALH. Dashed line, retraction landmark. Scale, 1 μm. Genotype: RN2-gal4, UAS-Chrimson::mVenus, UAS-cherry::zeus. (q) Quantification (two-way ANOVA): p<.008. N=3 animals, each with N=5 dendrites assessed. (r-s) Dendritic (myr::GFP) distribution of (r’-s’) microtubules (Cherry::Zeus) in (r-r’) controls and (s-s’) post-overexpression of Nrx-1 in MNs, 4 h ALH. (r-s’) Scale, 5 μm. Genotypes: (r-r’) RN2-gal4,UAS-myr::GFP x UAS-Cherry::Zeus,UAS-redstingerNLS (s-s’) RN2-gal4,UAS-myr::GFP x UAS-Cherry::Zeus,UAS-Nrx-1. (t-u) Quantification (one-way ANOVA) of dendrite volume (p<.009) or microtubule:dendrite volume (p<.0001). Denoted N= #animals. (v-w) Live imaging of stable microtubules (Cherry::Zeus+) in aCC/RP2 (v) control or (w) Nrx-1 overexpression dendrites. (v-w) Scale, 1 μm. Genotypes: (v) RN2-gal4,UAS-Cherry::Zeus x UAS-myr::GFP, (w) RN2-gal4, UAS-Cherry::Zeus x UAS-Nrx-1. Pseudocolor: stable (green), extending (pink), or retracting (blue) dendrites. (x) Quantification, Fisher’s Exact Test (two-sided, p<.04). Denoted N= #animals, each with N=10 dendrites assessed. (y) Summary. Throughout: Error bars, mean ± SD. Significance: *p<.05; **p<.01; ***p<.001; ****p<.0001; NS= not significant. ♦ used significance following two-way ANOVA when one-way and two-way are displayed together. Ns: biological replicates from 2 technical replicates.

Figure 5.. Critical period extension alters locomotor behavior.

(a) Experimental paradigm. Animals were reared at 30°C through 12 ± 2 h ALH to extend critical period plasticity, then shifted to 18°C until 44 h ALH (25° staging standard, see Methods). Genotypes: MNs: RN2-gal4,CQ2-gal4 x UAS-RNAi,TubP-gal80^ts; astrocytes: TubP-gal80^ts,25h07-gal4 x UAS-RNAi or control. (b) Five behavioral metrics assayed. (c-f) FIMTrack traces (imaged 1 minute, 4 Hz) from individual freely behaving larvae from control (N=38) and transient MN KD of lar (N=23) or nrx-1 (RNAi1 and RNAi2 each N=43). N= #animals. (g-k) Quantification and statistics (one-way ANOVA): p<.0001 for all genotypes, all metrics. (l-o) FIMTrack traces from individual freely behaving larvae from control (N=75) and transient astrocyte KD of chpf (N=83), nlg4 (N=80), or nlg2 (N=80). N= #animals. (p-t) Quantification and statistics (one-way ANOVA) relative to control ordered as in graph: (p) p<.13, p<.2, p<.02; (q) p<.55, p<.79, p<.96; (r) p<.02, p<.09, p<.007; (s) p<.02, p<.8, p<.86; (t) p<.05, p<.61, p<.99. KD of gat resulted in unrecoverable paralysis. Throughout: Error bars, mean ± SEM. Significance: *p<.05; **p<.01; ****p<.0001. Ns: biological replicates from 2 technical replicates.